Skip to main content

Full text of "BSTJ : Human Factors Engineering Studies of the Design and Use of Pushbutton Telephone Sets (Deininger, R.L.)"

See other formats


Human Factors Engineering Studies 

of the Design and Use of 

Pushbutton Telephone Sets 

By R. L. DEININGER 

(Manuscript received February 16, 1960) 

From the user's point of view, what are the desirable characteristics of 
pushbuttons for use in 500-type telephone sets? The studies reported bear 
on this question and also on questions of how people process information 
when keying telephone numbers. Four categories of design features were 
studied: key arrangement, force-displacement characteristics, button-top 
design and central office factors. The results indicate that considerable lati- 
tude exists for key set design in terms of user performance; however, the 
preference judgments are more selective. The studies also showed that the 
manner in which the person acquired and keyed the telephone number 
influenced performance appreciably. 

Technological progress in recent years has brought pushbutton signal- 
ing from the telephone set within sight of economic feasibility. What, 
from the user's viewpoint, are the desirable operating characteristics of 
the key set which should guide development and manufacture? And how 
do people process information when they key a telephone number? 

I. HUMAN FACTORS PROBLEM 

Specifically, we would like to know how pushbutton design influences 
user speed, accuracy and preference in keying telephone numbers. What 
design specifications will maximize these three quantities, and how 
critical is it to achieve these maxima? Also, what other factors influence 
user information processing in keying telephone numbers? For example: 
How does performance improve with practice? Does it matter how the 
number to be keyed is displayed? And are there systematic procedures 
that users follow in keying numbers? 

The design features are discussed in Section III and fall into four 
groups: (a) key arrangement, (b) force-displacement characteristics, (c) 

995 



996 THE BELL SYSTEM TECHNICAL JOURNAL, JULY 1960 

button top design and (d) central office considerations. Observations 
concerning other factors in keying behavior are presented in Section IV. 

II. EXPERIMENTAL APPROACH 

The number of possible key arrangements, force-displacement char- 
acteristics and button tops is very large — too large to be tested. A 
selection of characteristics was, therefore, made on the basis of prior 
knowledge, user expectation and broad engineering requirements, so 
that we could examine only the region around an expected maximum. 

In general, each series of test sessions extended over three to five days 
and compared variations of one or two characteristics, with all other 
characteristics being kept constant. At the end of the tests the preferred 
values of the individual characteristics were incorporated into a single 
key set. Evaluation of this set provided a check on the interactions of 
these individual choices and on how well they fitted together. 

It was recognized that people's keying experience with each set during 
the three to five sessions in a series would be limited compared to the 
years of practice they could get if pushbutton telephones became a 
reality. However, methodological studies showed that differences on a 
relative basis between key set designs appeared after a comparatively 
small amount of experience. 

A group of adjustable pushbutton telephone sets was used in the 
later studies in the series. To build a telephone set for every change in a 
characteristic would have been prohibitive in cost. Such changes were 
simulated by use of specially designed universal pushbutton switches 
(Fig. 1). Each adjustable telephone set contained ten universal switches 
mounted in an arrangement determined by the face plate employed. 

Typically, a sample of 10 to 15 employees at Bell Telephone Labora- 
tories, Murray Hill, N. J., was drawn randomly for each study. These 
people came to a laboratory test room and used two to five pushbutton 
telephone sets that differed in several characteristics, keying 10 to 15 
standard telephone numbers on a set each day. A different set was used 
for each daily session until all the sets in the study had been tried, the 
order of use being dictated by the design of the experiment. The number 
of individuals used in the later studies was selected to detect small 
differences among the average keying times, by considering the power of 
the analysis of variance tests involved (Ref. 1, p. 379 and 425). The 
error terms for these tests were reduced by removing the effects of 
practice and of differences among individuals in the analysis of variance. 

On the first day the subjects were told the procedure that would be 
followed and were asked to key accurately and quickly, as if they were 



HUMAN FACTORS STUDIES OF PUSHBUTTON TELEPHONES 



997 



CONTACT ASSEMBLY 



CONTACT CAM . 




PLUNGER STOP 



FRAME 



CROSS ROD 



PLUNGER 



SNAP SPRING 



CAM 



"--BUTTON 



Fig. 1 — The universal pushbutton switch. Changes in force-displacement 
characteristics were obtained by hand adjustments of the components. 

at home or in their office. After trying all conditions in the study, each 
person was asked for his preferences and for any suggestions he might 
care to offer. The use of employees in a laboratory-type study seemed 
justified for two reasons: All comparisons were made within the same 
study and presumably under constant conditions, and the field trials 
that followed the laboratory studies would serve to check the findings. 

When a person keyed a number, his performance was measured on 
the equipment shown diagrammatically in Fig. 2. Speed was measured in 
terms of the keying time, that is, the time interval from the electrical 
contact of the first key depression until the end of contact of the seventh. 
Keying accuracy was determined by automatically comparing the num- 
ber keyed with the number to be keyed. Provisions were made for ob- 
taining more detailed time measures, such as interdigital times and 
contact durations. 



III. THE DESIGN VARIABLES 

3.1 Arrangement of Keys 

The arrangement of the keys was specifically investigated in one study 
and then checked incidentally in several others. Although other methods 



998 



THE BELL SYSTEM TECHNICAL JOURNAL, JULY 1900 



VISUAL 

ERROR 

INDICATOR 



VISUAL KEYED 

NUMBER 

INDICATOR 



PUSH- 
BUTTON 
TELEPHONE 




K S.T!!£ C I (MAKE OF 1ST CONTACT TO % 
TIME I RELEASE OF 7TH CONTACT) 

CLOCK ■ 



PEN 
RECORDER 



CROSSBAR OR 

TRANSISTORIZED 

SWITCHING 

EQUIPMENT 



J 



PUNCHED 

TAPE 

TRANSLATOR 



PUNCHED 
. TAPE 
READER 



TAPE 
READER 
DRIVING 
CIRCUIT 



-NUMBER KEYED 



-+ 



NUMBER TO BE KEYED 



-A 



Fig. 2 — Schematic drawing of equipment for measuring and recording keying 
performance. 



were used to make the initial selection of arrangements for the study, 
performance tests were used to make the final choice because the other 
methods seemed inadequate. For example, a preliminary study by Lee 
and Snodgrass 2 showed that there was no significant relation between 
initial questionnaire preference and subsequent keying performance, 
and that 10 of the 20 subjects changed their preferences after using the 
pushbutton telephone sets. 

Sixteen different arrangements were selected for the first part of this 
study. These arrangements were compared by separating them into 
groups of three and having a different sample of six employees try all 
three arrangements during each of five sessions. The arrangements and 
how they were grouped for the study are shown in Fig. 3. 

In all the comparisons, only small differences were found in the keying 
times and errors, and the most preferred arrangements tended to be the 
best in terms of performance. Significant differences in keying times, 
errors or preferences were found in four of the six comparisons. Notice 
that the arrangement frequently found in ten-key adding machines (ar- 
rangement I-A, Fig. 3) was not the best of the first three arrangements 
compared. On the other hand, the same geometric configuration with a 
different numbering scheme (arrangement IV-A) was superior in keying 
performance when compared in Group IV. However, the performance 
differences between the two were small : arrangement I-A had an average 
keying time of 5.08 seconds, and arrangement IV-A had an average of 
4.92 seconds. 



GROUP I 




GROUP n 



n-c 




GROUP in 




m-c 



GROUP 15 



H-A* 




ffi-C 



GROUP 2 



Y-C 




GROUP 21 



-R* 



n-c 




•significantly shorter keying time 
t significantly lower error rate 



^SIGNIFICANTLY MORE PREFERRED 



Fig. 3 — The 16 arrangements used in the first study, grouped as they were 
compared. Two of the arrangements used earlier were compared again in the last 
group. Although not shown in the figure, the letter groups usually associated with 
the numbers on a telephone dial were also on the button tops. The tops were f inch 
square with ^-inch-high black letters and numbers on a white background. The 
circles shown were 41 inches in diameter. 



999 



1000 THE BELL SYSTEM TECHNICAL JOURNAL, JULY 1960 



ARRANGEMENT 




THREE-BY-THREE 
PLUS ONE 



KEYING TIME 
(SECONDS) 



PER CENT 
ERRORS 



RANKING 
FOR 



3RD 



RANKING 
AGAINST 



2ND 




TWO HORIZONTAL 
ROWS 



2.3 1ST (MOST) 



4TH 




TWO VERTICAL 
COLUMNS 



1.3 



5TH (LEAST) 



1ST (MOST) 




TFI EPHONE 



5.90 



2.0 



2ND 



5TH (LEAST) 




SPEEDOMETER 



5.97 



3.0 



4TH 



3RD 



Fig. 4 — The five arrangements compared in the second study. The specifica- 
tions listed in the caption of Fig. 3 apply also to this figure. 



HUMAN FACTORS STUDIES OF PUSHBUTTON TELEPHONES 1001 

The four arrangements found superior in their individual comparisons 
and the arrangement similar to the standard rotary dial (Fig. 4) were 
used in the next study. In this way, the fastest and most preferred ar- 
rangements were compared directly with the standard rotary dial ar- 
rangement. The rectangular arrangements had the buttons spaced with 
f inch between centers. The buttons in the circular arrangements were 
at the ends of lj-inoh radii and were separated by 30°, with the first 
and tenth buttons being separated by 90°. 

A new sample of 15 employees served in three replications of the 
same Latin square experimental design. 3 No significant differences were 
found among the keying times even though the study was designed to 
detect a one-half -second difference among the sets in nine out of ten 
instances, given the 95 per cent level of confidence and an error term 
estimated from the previous study. Similarly, no significant differences 
were found among the error rates. It was concluded that any of the 
five arrangements was acceptable, but that the arrangement with two 
vertical columns of keys should be avoided because it was disliked by 

many subjects. 

Although either rectangular or circular arrangements were found ac- 
ceptable, two of the rectangular arrangements in Fig. 4 offered certain 
engineering advantages and were studied further. A subsequent study 
showed that the buttons in the three-by-three-plus-one arrangement 
could be spaced f or f inch between centers without significant change 
in performance (5.56 seconds and 1.7 per cent errors versus 5.54 seconds 
and 2.5 per cent errors), although preferences indicated the larger 
spacing was more desirable. The buttons in the two horizontal rows 
arrangement could be spaced either f or H inch between centers with 
little effect on performance (5.33 seconds and 5.8 per cent errors versus 
5.63 seconds and 2.5 per cent errors), or on preference. 

3.2 Button Tops 

The button tops used initially in the adjustable pushbutton tele- 
phone sets were marginally acceptable from a legibility point of view 
(see Ref. 4, p. 24). Therefore, a study was conducted to assess the effects 
of larger letters and button tops on performance and preference. Five 
combinations of button-top size, letter size and location number size, 
and pushbutton arrangement were compared (see Fig. 5). Gorton con- 
densed letters and similar numerals were used in the sets, as were the 
stroke widths recommended by Baker and Grether 4 for the various 
conditions. 

Fifteen randomly selected employees served as subjects in three 5X5 



1002 



THE BELL SYSTEM TECHNICAL JOURNAL, JULY 1900 



BUTTON 
ARRANGEMENT 



SET AND 
SPACING 



BUTTON AND 
LETTERING SIZE 



REMARKS 



(3X3+1) 



SET NO. i: 

3/4 IN. 
BETWEEN 
BUTTON 
CENTERS 



H«-3/8-H 



3/8 

I 



9/64 



9/64 

r 



KEYING TIME: 6.35 SEC 
KEYING ERRORS: 7.1% 
2 VOTES FOR 
9 VOTES AGAINST 



SET NO. 2: 

3/4 IN. 
BETWEEN 
BUTTON 
CENTERS 




KEYING TIME: 5.83 SEC 
KEYING ERRORS: 1.3% 
1 VOTE FOR 
1 VOTE AGAINST 



(5-5-H) 



r*-i/2— H 




SET NO. 3: 

3/4 IN. 
BETWEEN 
BUTTON 
CENTERS 



SET NO. 4: 

3/4 IN. 

BETWEEN 

BUTTON 

CENTERS 

AND LETTERS 

ON PLATE 



KEYING TIME: 5.75 SEC 
KEYING ERRORS: 2.0% 
4 VOTES FOR 
1 VOTE AGAINST 



5/32 




KEYING TIME: 5.77 SEC 
KEYING ERRORS: 3.3% 
6 VOTES FOR 
4 VOTES AGAINST 



SET NO. 5: 
27/32 IN. 
BETWEEN 
BUTTON 
CENTERS 




KEYING TIME - . 6.07 SEC 
KEYING ERRORS: 5.3% 
2 VOTES FOR 
VOTES AGAINST 



Fig. 5 — The five conditions compared in the lettering study. Both the "votes 
for" and "votes against" total to 15, the number of subjects in the study. 



Latin squares, this number again being based on the power of the test 
considerations. 

Apparently, there is an optimal size for the button tops. Significant 
differences were found among the keying times and errors, with the 
smallest tops and lettering being the poorest. Keying with the middle- 
size tops was superior to that with the smallest tops, regardless of the 
arrangement of the keys or the location of the lettering. Keying with 
the large rectangular tops fell between that with the middle-size and 
that with the smallest-size tops, but did not differ significantly from 
either. As is apparent in Fig. 5, the differences among the keying times 



HUMAN FACTORS STUDIES OF PUSHBUTTON TELEPHONES 1003 

were relatively small, even though they were significant: The largest 
difference was 0.6 out of 6.0 seconds. The error rates differed significantly 
from one another; however, a large part of the x 2 was due to the ex- 
tremes of 1.3 and 7.1 per cent. 

The preferences followed the same pattern as the performances. Sub- 
jects stated their dislike of the smallest top and lettering and their 
preference for the middle size. Placing the letters on the plate rather 
than on the button top was controversial: Six individuals liked the 
letters off the top, but four others disliked the idea. 

The top of the pushbutton seemed to be important because it served 
as a display for the associated number and letters and as a target for 
the key-pressing response. Increasing the size of the top and lettering 
improved the display. Because the over-all size of the keyset was limited, 
the larger button tops ultimately required a reduction in the separation 
between adjacent tops, thereby impairing the qualities of the target. 

3.3 Force-Displacement Characteristics 

While the arrangements were being studied, three aspects of the force- 
displacement curves were under investigation. Prior to that time, pre- 
liminary studies showed that gross variations in the feel of the button 
had little effect on performance. However, these studies showed force 
in the neighborhood of 100 to 200 grams and displacements of about 
one-eighth inch were preferred to larger values. 

3.3.1 Force 

In the first study, the force required to depress the button was varied 
while the maximum displacement was held constant at one-sixteenth 
inch. Two conditions were used: medium-touch buttons and light-touch 
buttons (sec the dashed curves in Fig. 6). The forces referred to de- 
pended on the helical springs used in the button mechanisms, and Fig. 
6 shows static force-displacement curves due mainly to the springs. 

A sample of 24 employees used each condition for two consecutive 
days, half trying the medium-touch buttons first. Only small (0.04 
second and 0.8 per cent errors) and insignificant performance differences 
were found. The preference judgments were somewhat obscured by the 
fact that the subjects were not told the difference between the two sets 
until the end of the study and did not have the sets in front of them 
while they made their judgments. Fifteen persons failed to notice a 
difference, and meaningful preference information could not be obtained 
from them. However, of the nine who noticed a difference and had a 



1004 THE BELL SYSTEM TECHNICAL JOURNAL, JULY 1960 

500 



450 



400 



350 



w 300 
< 

Z 250 



200 



150 



50 





1 










jf 




' TRAVEL STUDY 




15 


/ 
































— 


10 










- 










J 


- 




J 


i 
/ 






- 


5 










- 


















100 ^gv 



1/32 1/16 



1/8 3/16 

DISPLACEMENT IN INCHES 



1/4 



Fig. 6 — The idealized force-displacement curves of the two conditions com- 
pared in the study of forces, and the three conditions compared in the study of 
button travel. These curves are based on essentially static measurements of the 
universal pushbutton switches. 

basis for their preference judgment, eight preferred the light-touch 
button and one had no preference. 



3.3.2 Travel 

Three conditions of maximum button displacement were compared in 
the next study. The force was held approximately constant, near 100 
grams, and maximum displacements of ■£$, % and y% inch were compared 
(see the solid curves in Fig. 6) . The point at which the electrical contact 
was made varied with maximum displacement, so that the contact was 
closed during 50 per cent of the travel. A new sample of 27 employees 
tried one condition a day in a cross-over experimental design 3 until all 
three had been used. 

The average keying times ranged from 5.67 to 5.86 seconds and the 



HUMAN FACTORS STUDIES OF PUSHBUTTON TELEPHONES 1005 

error rates from 3.9 to 4.1 per cent. No significant differences were 
found. When asked which condition they most preferred, 11 subjects 
said the ^-inch displacement, 9 the £-inch and 6 the ^-inch. Twelve 
voluntarily stated they did not like a particular condition: eight dis- 
liked the 3V inch, one the |-inch and three the iVinch. Taking into 
account both likes and dislikes, the smallest displacement appears 
controversial, the largest unpopular and the middle the most desirable. 

3.3.3 Feedback 

The next question studied was as follows: What would happen if 
additional auditory or kinesthetic-tactile feedback were added to a but- 
ton incorporation desirable values of force and travel? Letting the 
customer hear the voice-frequency switching signals might provide feed- 
back concerning the adequacy of the button pressings. Moreover, the 
addition of a slight snap action and a more distinctive bottoming to the 
pushbutton might improve performance. 

For this study, the buttons in one adjustable set had the force and 
displacement that were found desirable in the two previous studies. The 
buttons in a second set had the same travel, but had a slight snap action 
and a bottoming action that terminated movement abruptly and with 
an audible click. Each of these sets was used with and without the 
voice-frequency code signals, so that a total of four conditions were 
studied. Three 4X4 Latin squares were used with 12 employees, who 
tried one condition a day. 

For both performance and preference, the differences among the four 
conditions were small and insignificant. This study is interpreted as 
indicating that neither form of additional feedback was necessary in a 
button mechanism that had desirable force-displacement character- 
istics. 

3.3.4 Composites of Characteristics 

Because the studies reported thus far investigated one or two of the 
characteristics at a time, there remained the question whether per- 
formance and preference for composites of the characteristics could be 
predicted from the individual studies. This point was checked in the 
following study that compared three pushbutton telephone sets, each 
embodying a different composite of characteristics. 

The preferred characteristics isolated by the preceding studies were 
combined in one of the adjustable pushbutton sets. The other two sets 
in this study had been designed for various engineering and test purposes 



1006 THE BELL SYSTEM TECHNICAL JOURNAL, JULY 1960 

before the preferred values were isolated. When the characteristics of 
these two sets were assessed in terms of the individual human factors 
studies, it was found that they deviated from the most preferred values 
but were still largely within the range of desirable and acceptable values. 
The largest deviations were found in the force-displacement character- 
istics, but even these were not extreme. Thus, it was predicted that 

(a) small and insignificant performance differences would be found, but 

(b) large and significant preference differences would be found, particu- 
larly where the force-displacement characteristics of the button mecha- 
nism were concerned. 

Forty-five employees, none of whom had served in an earlier study, 
used one set each day to accord with 15 3 X 3 Latin squares. This 
sample size was selected to provide a more powerful test of the error 
rates than was given by the earlier studies and to permit more detailed 
analysis of the preference judgments. 

As predicted, the differences among the average keying times were 
small and insignificant. An average of 5.8 seconds was required to key a 
standard telephone number on the most preferred combination versus 
5.9 and 6.0 seconds on the other two. However, significant differences 
were found among the error rates because the electrical contacts on one 
set were out of adjustment. At least one error was made in keying 2.3 
per cent of the telephone numbers on the most preferred composite, in 
keying 2.0 per cent of the numbers on the second set and in keying 10 
per cent of the numbers on the third set. The additional 8 per cent errors 
were all of one type ; when the fault was corrected in later models, more 
normal error rates were obtained with that set. 

The large differences expected among the preference judgments for 
the three composites were found. In an interview after the last day of 
testing, the subjects were asked which set they preferred the most and 
which they preferred the least. Although there were no significant dif- 
ferences among the first place votes, the set with the most preferred 
characteristics was rated as "least liked" by significantly fewer people 
than either of the two other sets. The subjects were then asked which 
button feel they liked the most and which they liked the least. Again, 
the set with the most preferred characteristics came out ahead: not a 
single person disliked the feel of its mechanisms. 

In a sense, this study served to validate the procedure adopted in 
this series, since it showed that the desirable characteristics isolated in 
the individual studies could be combined into a superior set. Very likely, 
this was because much the same procedure was used throughout the 
entire series. Also, the fact that the characteristics of the other two sets 



HUMAN FACTORS STUDIES OF PUSHBUTTON TELEPHONES 1007 

were close to the characteristics investigated in the individual studies 
facilitated agreement between the predictions and the results. 

3.4 Central Office Factors 

How long the customer holds the pushbutton in contact and how long 
he pauses between consecutive key pressings are important in the design 
of central office switching equipment. Of primary concern are short 
contact durations and brief interdigital times. Some data regarding these 
two intervals were gathered in two studies. 

Fifteen employees keyed a total of 1500 seven-character all-numeral 
numbers on each of two pushbutton telephone sets in the first study. 
These sets were prototypes of operating equipment and differed in three 
characteristics: the arrangement of the keys, the force required to de- 
press the keys and the point at which the electrical contact was made 
as the button was being depressed. 

Keying was about as fast on one set as on the other; however, the 
contact durations were shorter and the interdigital times longer on one 
set than on the other (see Fig. 7). This differential proportioning of the 
time seems to reflect the relative location of the electrical contact in 
the travel of the button as indicated in the figure. The results of this 
study agree with the results of a preliminary study in which five proto- 
type key sets were studied. Thus, it may be possible to lengthen the 
contact duration by increasing the percentage of button travel during 
which contact is made. 

Very brief contact durations did occur; for example, durations equal 
to or less than 0.042 second occurred in 1 per cent of the key pressings. 
The interdigital times were longer, and 1 per cent of them fell at or be- 
low 0.090 second. The occurrence of short time values depended on the 
person doing the keying, particularly for the interdigital times, where 
most of the short values were due to one individual. Another source of 



PER CENT 
OF TRAVEL 



PER CENT 

OF KEYING 

TIME 



NO CONTACT 



NO CONTACT 



SET A 



J 



V///////. 



NO CONTACT 



NO CONTACT 



<%31%Z,, 



CONTACT 



Y/////////A 

Z36 °/oY//a 



SET B 



Fig. 7 — The effect of the location of the electrical contact on the percentage 
of keying time during which contact was made. 



1008 THE BELL SYSTEM TECHNICAL JOURNAL, JULY 1960 

short interdigital intervals was a repeated character in the number to 
be keyed. The interdigital time for repeated characters averaged 0.100 
second less than the times for nonrepeated characters. 

A second set of detailed measurements, taken during part of the 
study of composites, showed more clearly that brief contact durations 
and interdigital times were associated with fast keyers. The last ten 
keyings made by 30 of the 45 individuals using one of the prototype 
sets were recorded, equipment permitting. The correlations among the 
average contact duration, the average interdigital time and the average 
keying time for each subject were computed after the averages were 
transformed logarithmically to reduce the skewness of the distributions. 
The average keying time correlated 0.47 with the average contact dura- 
tion and 0.76 with the average interdigital interval. No relation was 
found between the contact duration and the interdigital time. The 
correlation coefficients indicate that fast keyers tend to have shorter 
contact durations, and definitely make shorter interdigital pauses. This 
latter point will be discussed in greater detail in a moment, for it seems 
related to the procedure that each subject adopted in keying numbers. 

IV. OBSERVATIONS ON KEYING BEHAVIOR 

When a person keys a telephone number he is processing information 
in a very literal sense. The standard telephone number contains 22.3 bits 
of information; on the basis of the key pressings, the central office 
switching equipment selects one line and completes the call. The physical 
characteristics of the pushbutton key set are one factor that influences 
the efficiency of the information processing. As was reported in Section 
III, the speed and accuracy of performance were affected by the arrange- 
ment of the keys and the size of the button tops. Moreover, if larger 
forces and longer displacements were studied, or if more powerful ex- 
periments were used, significant effects very likely would be found for 
the force-displacement characteristics as well. 

4.1 Additional Factors That Influence Performance 

The studies reported here and studies made by other groups at the 
Laboratories have pointed to additional factors that influence the 
efficiency of information processing. In the case of unfamiliar numbers, 
the manner in which the number is displayed is important. Keying un- 
familiar numbers from the pages of a telephone directory can increase 
the keying times by 75 per cent and the keying errors by 100 per cent 
in comparison with keying from a 3- X 5-inch file card on which only 



HUMAN FACTORS STUDIES OF PUSHBUTTON TELEPHONES 1009 

one number is typed. The amount of information to be processed is also 
a factor; for example, about 4 seconds more is required to key a seven- 
character telephone number than a four-character one. Familiar numbers 
and numbers with repeated characters or simple sequences of digits are 
keyed quickly and accurately. 

Experience with pushbutton keying facilitated the information process- 
ing. Keying became faster as people used the key sets day after day. 
However, performance seemed to improve at about the same rate 
whether they used the same set each day, the same two or three sets 
each day, or a different set each day. Some of this improvement was due 
to increased familiarity with the test room and procedure, but similar 
improvements with practice appear in field trials, indicating that ex- 
perience in using pushbuttons is important. In the case of keying ac- 
curacy, the low error rates make detection of any learning trends dif- 
ficult. 

Perhaps the most important factor in the information processing is 
the individual himself. Some people keyed seven-character telephone 
numbers in less than two seconds, requiring about 2.7 seconds on the 
average, and others required as high as 12.4 seconds on the average to 
key the same numbers. These differences were consistent from day to 
day and from telephone to telephone, as can be gathered from the fol- 
lowing rank-order correlation coefficients: There was a 0.8 correlation 
between pushbutton keying times on day 1 and pushbutton keying 
times on days 12 through 14, a 0.7 correlation between pretest rotary 
dialing times and pushbutton keying times on day 1 and 0.6 correlation 
between pretest dialing times and pushbutton keying times on days 12 
through 14. The keying times on two different pushbutton sets were 
more highly correlated than rotary dialing times and pushbutton keying 
times. 

4.2 The Importance of the Keying Process 

The large differences among the average keying times for individuals 
are largely due to the ways in which people acquire and key telephone 
numbers. It is a question whether they memorize the entire number and 
key it without referring back to the display, or whether they memorize 
and key the first part of the number and then refer back to the display 
to memorize the remainder for keying. Detailed analysis of the inter- 
digital times shows that the intervals for the fastest subject vary only 
slightly (see Fig. 8). The interdigital times for average-speed keyers 
show a distinct increase between the third and fourth characters. This 
accounts for the fact that the average keying time is more highly cor- 



1010 THE BELL SYSTEM TECHNICAL JOURNAL, JULY 1960 



1.50 



? 1.00 

5 



W 0.75 




3 4 5 

POSITION OF CHARACTER 



Fig. 8 — Average interdigital intervals for individual subjects keying seven- 
character numbers. The interdigital intervals were measured from the breaking 
of one electrical contact to the making of the next contact. (These results were 
obtained for all-numeral dialing; data for letter-and-numeral dialing are similar 
but less definitive than those shown in the figure.) 

related with the interdigital time than with the contact duration. These 
findings corroborate the statements made by people about how they 
key telephone numbers. The fastest say that they memorize and key 
the entire number without referring back to the display, whereas the 
average to slow keyers say that they refer back in the middle of keying 
the number. 

4.3 Why Refer Back? 

Neither method of keying seems more accurate than the other. One 
might predict that keying without referral would be less accurate and 
therefore that fast keyers would make more errors than slow keyers. 
However, no relation between keying time and errors could be detected 



HUMAN FACTORS STUDIES OF PUSHBUTTON TELEPHONES 1011 

among the studies reported, regardless of whether they were considered 
singularly or as a group for analysis. The absence of a correlation be- 
tween time and errors in the present studies must be interpreted cau- 
tiously, since these studies were not designed to detect such relations. 

On the other hand, referral does serve an important purpose, as indi- 
cated by the following study. A sample of six employees was asked to 
key telephone numbers with referral during two sessions and without 
referral during two other sessions. When keying without referral, the 
individuals were asked to memorize the entire number, turn over the 
display card, and key the number from memory. A total of 156 standard 
telephone numbers were keyed under each set of instructions. Some sub- 
jects found it very difficult to key telephone numbers without referral. 
These people said they usually dialed unfamiliar telephone numbers by 
breaking them into two or more parts. The difficulty they experienced 
was reflected in their error rates. There was a drop in keying accuracy 
as more and more of the characters in the number were keyed, and this 
drop was due primarily to those who said they usually dialed with 
referral (see Fig. 9). Apparently, if a person habitually keys with referral, 



< 0.95 

X 

o 



z 

C 0.90 



0.85 



0.75 



^n 
^ 


» 




„»«l 


w«. 






.4--. 












»••*• "V-, 
\'^^ 






~~H 


i""" 


WITH 
REFERRAL 


""""l 


>____. 


— — — -A 
































































































( i 












without\ 
referral 





















































































3 4 5 

POSITION OF CHARACTER IN NUMBER 



Fig. 9 — Accuracy of keying as a function of position of character for two 
methods of keying. The number display was always present during keying with 
referral, but it was removed before keying started for keying without referral. 



1012 THE BELL SYSTEM TECHNICAL JOURNAL, JULY 1960 

he does so to increase the accuracy of keying the latter characters in the 
telephone number, even though his over-all keying accuracy is not ap- 
preciably greater than that of a person who usually keys without referral. 

V. CONCLUSIONS 

Regarding the design of pushbutton telephone key sets that are fast, 
accurate and convenient to use, the following statements can be made: 

i. The operating characteristics of the key sets significantly influence 
both keying performance and user preference. 

ii. In terms of keying performance there exists a rather broad region 
of desirable values for the operating characteristics. Thus, there is 
latitude for telephone key sets and, so long as the characteristics remain 
in the region of desirable values, little deterioration in keying perform- 
ance will be found. 

iii. Considerably less latitude exists if preferences are considered, 
particularly in the case of the force-displacement characteristics. Typi- 
cally, subjects preferred a smooth and quietly operating button with a 
light touch and a moderate travel. 

Other factors that influence keying performance are practice, number 
length and display media, and familiarity with the telephone number. 
On the other hand, the most important factor influencing performance 
observed in these studies was the manner in which the subject acquired 
and keyed the number. A person who memorized the entire number and 
keyed it without referring back to the display could key a number in 
less than two seconds. However, a person who memorized part of the 
number, keyed it and then referred back to the display to memorize 
and key the remainder of the number could require more than 12 seconds 
to key the same number. 

VI. ACKNOWLEDGMENTS 

Many individuals contributed to the studies reported. In this group 
were Miss N. L. Bowles, P. D. Bricker, Mrs. S. L. Ferguson, 0. 0. 
Gruenz, Miss V. A. Hansen, Miss M. J. Kellogg, Miss E. T. Leddy, 
W. A. Lee, S. E. Michaels, W. A. Munson, R. R. Riesz, Mrs. S. B. 
Sheppard, Miss J. G. Snodgrass, Miss C. M. Steadier and C. H. Stumer. 

The author would like to thank J. E. Karlin for his advice in conduct- 
ing the latter phases of the program and in writing this report. 

REFERENCES 

1. Hald, A., Statistical Theory with Engineering Applications, John Wiley and 

Sons, New York, 1952. 

2. Lee, W. A. and Snodgrass, J. G., On the Relation Between Numbering Pref- 

erences and Performance on a Ten-Button Keyboard, Amer. Psychol., 13, 
1958, p. 425. 

3. Federer, W. T., Experimental Design, Macmillan, New York, 1955. 

4. Baker. C. A. and Grether, W. F., Visual Presentation of Information. Wrieht